Statistical strategies for event rate comparisons in dental studies

Area of Science:

• Dental Research
• Biostatistics
• Epidemiology

Background:

• Analyzing associations between dental variables and disease indicators at the tooth or surface level presents challenges.
• Standard chi-square tests are insufficient due to the inherent correlation of teeth within an individual.
• Stratification adjustments, especially for tooth-level factors, further complicate traditional analyses.

Purpose of the Study:

• To introduce statistical methods for analyzing correlated dental data at the tooth or surface level.
• To provide a framework for controlling intra-patient correlation in dental research.
• To enable accurate estimation of measures of association adjusted for within-patient clustering.

Main Methods:

• A survey sampling approach is proposed, treating patients as one-stage cluster samples.
• The design-based covariance matrix for cell counts in contingency tables is derived using statistical software for large sample surveys.
• Taylor series methods are employed to approximate variances for various measures of association.

Main Results:

• The proposed methods allow for event rate comparisons using odds ratios, relative risks, and risk differences.
• Adjustments for intra-patient correlation are successfully incorporated into the analysis of association measures.
• The approach facilitates summary measures of association across different strata, such as tooth type.

Conclusions:

• The survey sampling approach effectively addresses the correlation of dental sites within patients.
• This methodology enhances the accuracy of association measures in dental research involving tooth- or surface-level data.
• The findings support more robust statistical analyses for understanding dental disease indicators and explanatory variables.